NuGNN: A Graph Neural Network for Nuclear Reaction Network Equations

Abstract

Nuclear reaction networks are a major computational bottleneck in astrophysical simulations when large isotope sets are required, because of the stiffness of the network equations and the repeated calls to Jacobian-based solvers required by implicit methods. In this work, we develop a deep learning surrogate solver for a large 690 isotope nuclear reaction network over a broad range of thermodynamic conditions relevant to Type I X-ray bursts using a graph neural network, NuGNN. Unlike conventional fully connected or convolutional neural networks, NuGNN directly reflects the structure of the reaction network through heterogeneous isotope and reaction nodes and message-passing along reaction connections. The model is trained on data spanning many orders of magnitude in stellar temperature and density, as well as in simulation time-step size. We compare NuGNN with a Res-U-Net and fully connected neural network, and find that NuGNN consistently achieves significantly better accuracy with errors of only a few percent. More importantly, when implemented in the network evolution code in place of the original solver, NuGNN successfully reproduces the final abundance patterns, whereas the other architectures fail to do so. We also show that the trained model has a potential to improve computational speed. These results show that graph neural networks can provide a robust and promising framework for accurate surrogate modeling of large nuclear reaction networks.

A Hybrid Neural Network Emulator for M1 Neutrino Reaction Rates in Binary Neutron Star Merger Simulations
Yi Qiu, David Radice, Raffaele Ferrari +1
arXiv20 Oct 2026PreprintarXiv licenceFull textHigh-energy astrophysicsarXiv
Tighter dark-matter constraints from the projected mass method: A neural-network-enhanced method for galaxy groups and clusters
Yinbo Huang, David Benisty
Astronomy & Astrophysics4 Nov 2026Peer-reviewedPublisher termsSolar & stellar physicsCrossref
Detection of metal absorption lines in quasar spectra. A neural network approach using U-Net
Elena Sofia Mangola, Francesco Pistis, Michele Fumagalli +7
Astronomy & Astrophysics2 Oct 2026Peer-reviewedPublisher termsGalaxiesCrossref
Photometric Redshift Estimation with Hybrid Quantum Neural Network
Giuseppe Sarracino, Vincenzo Fabrizio Cardone, Irene Graziotti +13
arXiv26 Oct 2026PreprintarXiv licenceFull textGalaxies Quantum physics & quantum informationarXiv
NERV: Neural-network Enhanced Reconstruction of the UniVerse with Application to Baryon Acoustic Oscillations in the BOSS DR12 Galaxy Sample
Shi-Hui Zang, Hong-Ming Zhu, Tian-Xiang Mao +1
arXiv14 Sep 2026PreprintarXiv licenceFull textCosmologyarXiv

Cite

BibTeX RIS

BibTeX

@article{kim2026nugnn,
  title = {{NuGNN: A Graph Neural Network for Nuclear Reaction Network Equations}},
  author = {C. H. Kim and K. Y. Chae and S. Ko and M. R. Mumpower and M. S. Smith},
  journal = {The Astrophysical Journal Supplement Series},
  year = {2026},
  month = nov,
  doi = {10.3847/1538-4365/aeaa46},
  url = {https://iopscience.iop.org/article/10.3847/1538-4365/aeaa46},
  note = {CC BY 4.0},
}

RIS

TY  - JOUR
TI  - NuGNN: A Graph Neural Network for Nuclear Reaction Network Equations
AU  - C. H. Kim
AU  - K. Y. Chae
AU  - S. Ko
AU  - M. R. Mumpower
AU  - M. S. Smith
PY  - 2026
DA  - 2026/11/04
JO  - The Astrophysical Journal Supplement Series
DO  - 10.3847/1538-4365/aeaa46
UR  - https://iopscience.iop.org/article/10.3847/1538-4365/aeaa46
AB  - Nuclear reaction networks are a major computational bottleneck in astrophysical simulations when large isotope sets are required, because of the stiffness of the network equations and the repeated calls to Jacobian-based solvers required by implicit methods. In this work, we develop a deep learning surrogate solver for a large 690 isotope nuclear reaction network over a broad range of thermodynamic conditions relevant to Type I X-ray bursts using a graph neural network, NuGNN. Unlike conventional fully connected or convolutional neural networks, NuGNN directly reflects the structure of the reaction network through heterogeneous isotope and reaction nodes and message-passing along reaction connections. The model is trained on data spanning many orders of magnitude in stellar temperature and density, as well as in simulation time-step size. We compare NuGNN with a Res-U-Net and fully connected neural network, and find that NuGNN consistently achieves significantly better accuracy with errors of only a few percent. More importantly, when implemented in the network evolution code in place of the original solver, NuGNN successfully reproduces the final abundance patterns, whereas the other architectures fail to do so. We also show that the trained model has a potential to improve computational speed. These results show that graph neural networks can provide a robust and promising framework for accurate surrogate modeling of large nuclear reaction networks.
SN  - 0067-0049
SN  - 1538-4365
ER  -

Abstract

Related

Cite